Filetype PDF | Posted on 21 Sep 2022 | 3 years ago
Authentication
digital resources
82x Filetype PDF File size 0.77 MB Source: www.iaras.org
International Journal of Education and Learning Systems
Mona Hafez Mahmoud http://iaras.org/iaras/journals/ijels
A Multiagents based Intelligent Tutoring System for teaching Arabic
Grammar
MONA HAFEZ MAHMOUD
Informatics Research Department
Electronic Research Institute
El-Tahrir st., Giza
EGYPT
Monah1957@hotmail.com http:\\www.eri.sci.eg
Abstract
Intelligent agent has been around for years, but the actual implementation is still in its early stages. This
research is a scientific mix between two big topics of Artificial Intelligence. These topics are: the Intelligent
Agents and the Intelligent Tutoring System. An Intelligent Agent is a set of independent software tools that
are linked with other applications and database software running within a computer environment. The
primary function of an Intelligent Agent is to help a user (client) to better interact with a computer
application. It is assumed that artificial intelligence (AI) is involved and certain degree of autonomous
problem solving ability is presented in agent-based technology systems[1].
Intelligent Tutoring Systems (ITSs) simulates the one-to one human tutor for delivering knowledge
interactively instead of using books and the traditional learning environment. To come up with the most
learning outcomes, ITSs have incorporated several techniques such as: error identification and correction,
and building consistent explanations through integrating techniques of cognitive science and Artificial
Intelligence. Different tutoring systems have been implemented to cover different subjects and languages
such as: English, Arabic, Chinese, German and many others [2]. In this research ITS is covering grammar of
Arabic language. The global structure of ITS consists of mainly four modules: a pedagogic module, a
question selector module, an expert module and a student module in addition to a user interface module. But
in this system we didn’t need an expert module because we used Constraints Based Model (CBM)
technology (that will be explained below). This work is implemented under a project that is called
AG_TUTOR (Arabic Grammar tutor). This project simulates the behavior of instructors and students and
the relations between them in teaching the course of the Arabic Grammar of the fourth grade of the
elementary stage in Egypt. In this system the technology of Intelligent Agents is used. This research
concentrates on the Intelligent Agents part of AG_TUTOR.
Keywords Terms — Artificial Intelligence and education, Intelligent Tutoring System, Intelligent Agents,
Multi-Agents systems, knowledge base, domain knowledge.
1. Introduction: thus must communicate), and mobile agents
In computer science, an intelligent agent is a (agents that can relocate their execution onto
computer program that acts for a user or other different processors).[3]
program in a relationship of agency. In
particular, exhibiting some aspect of artificial
intelligence such as learning and reasoning are
related and derived concepts include intelligent
agents. There are many types of intelligent
agents: autonomous agents (capable of
modifying the way in which they achieve their
objectives), distributed agents (being executed
on physically distinct computers), multi-agent
systems (distributed agents that do not have the Fig. (1) Simple autonomous agent
capabilities to achieve an objective alone and
ISSN: 2367-8933 52 Volume 3, 2018
International Journal of Education and Learning Systems
Mona Hafez Mahmoud http://iaras.org/iaras/journals/ijels
As shown in Fig. (1), an intelligent agent (IA) is components for it are Task-Skill-Principle
an autonomous entity which observes through Editor, Exercise Editor, Student Model Editor,
sensors and acts upon an environment using and Tutor Behavior Editor. Each of these editors
actuators and directs its activity towards has their own specific functionality. An
achieving goals. Intelligent agents may also learn instructional agent is used to carry out
or use knowledge to achieve their goals. They instructional goals. It used Bayesian inference to
may be very simple or very complex. [4] incorporate student modeling strategies.[6]
A simple agent program can be defined MathTutor: it is a multi-agent ITS building
mathematically as an agent function which maps tool. Math Tutor integrates different formalisms
every possible percepts sequence to a possible in order to facilitate the teacher task of
action the agent can perform to a coefficient, developing the contents of a tutorial system and
feedback element, function or constant that at the same time to provide adaptively and
affects eventual actions: [4] flexibility in the presentation. Multi-Agent
Systems (MAS) technology have been of great
Intelligent Tutoring Systems (ITSs) are complex help in reducing the distance between ideal
computer programs that manage various systems and what can really be implemented,
heterogeneous types of knowledge ranging from because it allows to simplify the modeling and
domain to pedagogical knowledge. ITS typically structuring tasks through the distribution, among
consists of: different agents of the domain and student
1. The Pedagogic module, which designs and models. The proposed tool is based on a
regulates instructional interactions with the conceptual model, called MATHEMA that
students; provides a content-directed methodology for
2. The Question Selector Module, which selects planning the domain exposition and teaching
a question from a question bank, presents it to strategies. [8] [7]
the student and gets his response; Popular Tetris computer game: In this game a
3. The Expert module, which simulates human user must try to make a wall out of irregularly
experts in decision making or the instructor in shaped falling blocks. The agent in the game
education to get the correct answer of the takes the part of the user, who must control
question that presented to the student. (we didn’t where the blocks fall. Using traditional AI
need this module because we used CBM that will techniques would require representing
be explained later). knowledge about the game and the role of the
4. The Student Module, which is a dynamic user in terms of symbolic data structures such as
representation of the students current state of rules, and so on. This approach would be entirely
knowledge; unrealistic for a game like Tetris, which has hard
5. The user Interface module, which controls real-time constraints. Wavish and colleagues
interaction between the student and the system thus use an alternative reactive agent model
[5]. called RTA (Real Time Able). In this approach,
agents are programmed in terms of behaviors
which are simple structures. These agents are
2. Related work: loosely resemble rules but do not require
Here are some examples of systems that use the complex symbolic reasoning.[9]
intelligent agent's technology: UCEgo is a natural-language system that helps
FlexiTrainer: it is an authoring framework the user to solve problems in using the UNIX
that allowed a fast design of pedagogically rich operating system. It is the intelligent agent
and performance-oriented learning environments component of UC (UNIX Consultant). UCEgo
with tradition content and tutoring strategies. provides UC with its own goals and plans by
This authoring tool specifies a dynamic behavior adopting different goals in different situations. It
of tutoring agents that interact to deliver creates and executes different plans, enabling it
instruction. FlexiTrainer has been used to to interact intelligently with the user. Also, it
develop an ITS for training helicopter pilots in adopts goals from its themes, sub-goals during
flying skills. It consists of two components: the planning, and meta-goals for dealing with goal
authoring tools and the routine engine. Core interactions. It also considers goals when it
ISSN: 2367-8933 53 Volume 3, 2018
International Journal of Education and Learning Systems
Mona Hafez Mahmoud http://iaras.org/iaras/journals/ijels
notices that the user either lacks necessary attached to exactly one course topic or sub-topic.
knowledge or has incorrect beliefs. In these So, a method based on course structure is used. It
cases, UCEgo plans to volunteer information or uses a structure called a prerequisite structure
correct the user’s misconception, as that defines each course topic which other topics
appropriate.[10] the student must already know before proceeding
The organization of this paper will be as follows: further.
section 3 presents the Domain Knowledge,
section 4 presents the Knowledge Base while 4. Knowledge base:
section 5 presents the general structure of A knowledge base (KB) is a technology used to
AG_TUTOR, section 6 shows the Multi Agent store complex structured and unstructured
System in AG_TUTOR, Finally, section 7 information used by a computer system for
concludes the whole work. artificial intelligence domain. A knowledge-
based system consists of a knowledge-base that
represents facts about the world and an inference
3. Domain knowledge: engine that can reason about those facts and use
Domain knowledge in artificial intelligence is rules and other forms of logic to deduce new
the knowledge about the environment in which facts or highlight inconsistencies. [13]
the target system operates. The domain model Two relational databases are used in
organizes the course structure, its various AG_TUTOR which are implemented using
components and the relationship among the mySQL. One of them is considered the "lexicon"
components. This model mainly deals with the of the system which contains all the words
what-to-teach part of an ITS.[11] A domain (nouns, verbs, particles) of the exercises that will
model is created in order to represent the be represented to the student and their features.
vocabulary and key concepts of the problem The other one includes a bank of questions. Also,
domain. It also identifies the relationships among it includes all constraints, skills of the student,
all the entities within the scope of the problem feedback messages and all information about the
domain, and commonly identifies their attributes. student and his knowledge.
An important advantage of a domain model is
that it describes the scope of the problem
domain.[12] The adopted domain is the 5. The general structure of
curriculum of the grammar of the Arabic AG_TUTOR:
language of the fourth grade of the elementary In our proposed system, the system will deal
schools in Egypt. The knowledge of this with group of intelligent agents or multi-agent
curriculum is acquired from the Arabic instructor systems that deal with the modules specially
transcripts. Each lesson of the curriculum is student model. These agents are used for
considered a concept. Each concept may have learning and reasoning, also for modifying the
sub-concepts. Specifically, they cover the learning strategy, so every student can have a
following concepts and sub-concepts: different learning strategy according to his
individual problem diagnosis, or according to the
st nd
demonstrative nouns, pronouns (1 pronoun, 2 type of the student ( such as talent, smart, shy,
rd
pronoun, 3 pronoun), speech (Nouns, verbs, slow or fast in understanding and so on......).
particles), dual, plural, nominal and verbal Each module of the Educational system will deal
sentence, Interrogative and Negative tools and with one or more of intelligent agents. As will be
agreement of verb with the subject in gender. seen later, each one or more of the system agents
will represent an environment in the system such
ريمض ,ملكتملا ريمض) رئامضلا,ةراشلإا ءامسأ as the student, the teacher, the learning process
فورحلا ،لاعفلأا ،ءامسلأا) ملاكلا ,(بئاغلا ريمض.،بطاخملا and relations between them. Fig.(2) illustrates
ىفنلا تاودأ , ةيمسلإا ةلمجلا ,ةيلعفلا ةلمجلا،عمجلا ,ىنثملا ،( the structure of an ITS.
.ثينأتلاو ريكذتلا ىف لعافلا عم لعفلا قفاوت ، ماھفتسلإاو
Pedagogic Interface
Our course domain is richly articulated in topics module module
and subtopics (or concepts and sub-concepts). It Knowledge base
is required that each question in the domain is Question
Domain knowledge selector
module
Group of intelligent
agents Student
module
ISSN: 2367-8933 54 Volume 3, 2018
International Journal of Education and Learning Systems
Mona Hafez Mahmoud http://iaras.org/iaras/journals/ijels
of the question from the data base and presenting
to the student.
5.3. Student module:
Fig. (2) The structure of A Constraint Based Modeling (CBM) system is
adopted in implementing this module. The
AG_TUTOR concept of state constraints was invented to solve
5.1. Pedagogic module: a deep puzzle about skill acquisition: Human
The Pedagogic Module is a computer tutor that beings can catch themselves making errors. This
mimics the course patterns and educational ability forces a distinction between generative
tactics of a real human tutor [2]. It is the and evaluative knowledge. The function of
instructional module that designs and regulates generative knowledge (e.g., a rule set) is to
the instructor transcripts. The function of the produce actions according to the current
tutoring module is essentially to perform problem. And, the function of evaluative
continuous assessment of the student, and knowledge (a set of constraints) is to evaluate
thereby interact with the expert module to action outcomes as desirable or undesirable.
prescribe further action. [13][14] Different learning theories have different
In AG_TUTOR, this module is representing the implications for the design of ITSs. The state
concepts in a very attractive interface with high constraint theory suggests that the knowledge
quality of graphic, animation and sound. Also, it base of a constraint-based tutoring system should
represents group of examples for each concepts. contain the constraints that the student would
At the case the student’s answer is wrong; the have. Hence, such a tutoring system plays the
system will take the student back to the tutoring role of an amplified evaluative knowledge base.
module to give him more explanation about the [16]
concept. This module deals with Teaching The CBM is represented by a set of constraints;
Assistance Agents (that we will talk about later) each constraint represents a pedagogically
that helps in the teaching process for all the significant state [17]. The basic definition of a
lessons in our curriculum. constraint is formalized as:
<
5.2. Question selector module: Satisfaction condition >
The main goal of the question selector module is
to select a question randomly according to the Where the relevant condition is the condition
lesson that the student selects, display it to him that represents situations where constraint
and give him the chance to answer. [15] applies, satisfaction condition is the condition
This module drives these questions from the that has to be true in order for the constraint to be
question bank in the data base. The question satisfied, feedback actions is the action
bank contains a huge number of questions. The associated with the violation of the constraint.
bank is divided into many groups of questions as Constraint-based modeling has many benefits
a group for each lesson mainly: mainly:
1. Multiple Choices Questions (MCQ) Decreasing the time required to build an ITS
2. Match the related correct sentence by providing detailed and specific feedback
3. Press on something (like ريمض وأ ةراشإ مسا associated with the constraints.
بطاخم) The incorrect answers are implicitly
4. Fill in the space with the correct answer from implemented in the constraints, so no need to
the brackets implement them in the domain model in form of
5. Get out a verb, a noun, or a particle or …….. buggy-rules like model tracing.
6. Parse a sentence Changing any constraint in CBM has no
7. Reorder a nominal sentence to be a verbal effect on the other constraints at all.
sentence and vice versa. No need to the Expert Module to get the
8. Generate the plural, double or single of a noun correct answer.
In this module, Constraints and Hints Agent (that For modeling the student knowledge or skill in
will be explained later) is helping in the selection the linguistic domain, the constraint form is
modified to be as following:
ISSN: 2367-8933 55 Volume 3, 2018
no reviews yet
Please Login to review.
